Dr. Joseph Felsenstein of the University of Washington proposes a 5-year study of methods for constructing likelihoods of DNA sequences under the influence of recombination and selection. This new work builds on his previous likelihood evaluations for phylogenetic trees. Dr. Felsenstein's specific objective is to develop methods for analyzing data from several sequences of the same region from a single population with the special interest being to extend ML methods to accommodate recombination among sites. These methods will allow inferences to be made about the patterns of natural selection at specific sites, and will allow estimation of population parameters such as population size or migration rate. The feature of consideration of recombination between adjacent nucleotide sites will be allowed as a means of adding to the realism of the analyses. Although the method of maximum likelihood offers the best statistical framework on which to base inferences from data, there are considerable computational difficulties in evaluating likelihood functions. The problem is that sums must be taken over all possible gene trees linking any sample of sequences. Recombination increases this number. Modern computing methods, including applications of the Metropolis algorithm, allow progress to be made by sampling some of the possible trees in a way that depends on their contribution to the overall likelihood. The necessary computer programs are time-consuming to construct and check, and Dr. Felsenstein will make them available without charge over the Internet. He has previously made available his phylogeny construction package PHYLIP.
Showing the most recent 10 out of 16 publications
